Contrastive Explanations

Repo for the paper "Cold-Start Promotional Sales Forecasting through Gradient Boosted-based Contrastive Explanations", publicly available here

Experiments in the paper available as notebooks

• Experiment 1 Variable Importance
• Experiment 2 Surrogate model

Benchmark against other methods on a public dataset

Boston Housing: contrastiveRegressor/Benchmark_BostonHousing.ipynb

Contact

Any questions/remarks, feel free to drop me a line carlos.aguilar.palacios@gmail.com